Conversation: 02
19 Feb 2026 11:15h - 11:30h
Conversation: 02
Session at a glance
Summary
The discussion featured Amit Zavery, President and Chief Product Officer of ServiceNow, speaking with CNBC’s Arjun Kharpal about the role of trust as foundational infrastructure in enterprise AI adoption. Zavery emphasized that without trust, safety, and visibility into AI systems, enterprises cannot effectively implement AI due to auditing, compliance, and operational transparency requirements. He outlined a multi-faceted approach to building trust, starting with the human element through employee reskilling and cultural transformation to help workers understand AI’s value in making their jobs more efficient rather than replacing them.
ServiceNow has successfully implemented this approach across departments, seeing significant adoption increases as employees recognize AI’s ability to eliminate repetitive, “skull-crushing” tasks while freeing them for higher-value work. Zavery argued that companies using AI as an excuse for layoffs are missing the opportunity to reinvest savings into new market segments and capabilities. A critical component of enterprise AI adoption is security, particularly for agentic AI systems that can change roles dynamically. ServiceNow has invested heavily in security infrastructure, including acquiring companies like Vesa to manage non-human identity access, which contributed to a 55-fold increase in agentic workflow adoption.
Regarding market adoption, Zavery noted that while initial expectations were overly optimistic, thoughtful implementation with proper security controls has accelerated genuine enterprise adoption. He disagreed with predictions that AI would replace 50% of enterprise software, instead arguing that AI must be integrated as a foundational platform element, with successful companies building 90% of their IP around contextual understanding while leveraging foundational models for only 5-10%. Looking ahead, Zavery predicted continued focus on security, regulatory compliance, and the emergence of physical AI in operational technology and manufacturing environments.
Keypoints
Major Discussion Points:
– Trust as foundational infrastructure for enterprise AI adoption – Zavery emphasized that without trust, safety, and visibility into AI systems, enterprises cannot effectively deploy AI due to auditing, compliance, and operational requirements
– Human workforce transformation and reskilling strategies – Discussion covered how organizations can manage employee concerns about job displacement through retraining, cultural shifts, and demonstrating AI’s value in eliminating repetitive tasks while creating opportunities for higher-value work
– Security and safety challenges in agentic AI systems – Extensive focus on cybersecurity vulnerabilities, the need for proper access controls, identity management for AI agents, and how security must be built into AI products rather than added as an afterthought
– Enterprise AI adoption reality versus expectations – Zavery provided a realistic assessment of how enterprise AI adoption has been more thoughtful and slower than initially hyped, but is accelerating as security and control solutions mature
– Future of software industry and AI integration – Discussion of whether AI will replace traditional software (responding to claims that 50% of enterprise software could shift to AI) and the evolution toward physical AI in operational technology environments
Overall Purpose:
The discussion aimed to provide a realistic, enterprise-focused perspective on AI adoption challenges and opportunities, drawing from ServiceNow’s experience with millions of enterprise deployments to address both the promise and practical realities of implementing AI in business environments.
Overall Tone:
The tone was consistently pragmatic and measured throughout. Zavery maintained a balanced perspective, acknowledging both AI’s transformative potential and the legitimate concerns around security, workforce impact, and implementation challenges. The conversation remained professional and informative, with Zavery providing concrete examples from ServiceNow’s experience rather than engaging in typical tech industry hype.
Speakers
– Moderator: Role/Title: Event moderator; Area of expertise: Not specified
– Amit Zavery: Role/Title: President and Chief Product Officer, ServiceNow; Area of expertise: Enterprise software and AI, agentic AI implementation in enterprise workflows
– Arjun Kharpal: Role/Title: Senior Tech Correspondent, CNBC; Area of expertise: Technology journalism and reporting
Additional speakers:
None – all speakers in the transcript were included in the provided speakers names list.
Full session report
The discussion between Amit Zavery, President and Chief Product Officer of ServiceNow, and CNBC’s Arjun Kharpal explored enterprise AI adoption challenges and opportunities, focusing on ServiceNow’s experience with enterprise deployments. The conversation centered on Zavery’s assertion that “trust is the new infrastructure in this age of AI” and the practical realities of implementing AI in enterprise environments.
Trust as New Infrastructure
Zavery’s central thesis positioned trust as foundational to AI adoption, arguing that enterprises need safety, understanding, and visibility into AI systems to meet basic business requirements around auditing and compliance. Without these elements, he suggested, enterprises cannot effectively implement AI because they cannot operate systems they cannot audit or understand. This creates a fundamental barrier where AI systems functioning as “black boxes” conflict with core enterprise governance needs.
Workforce Transformation Through Cultural Change
The discussion revealed ServiceNow’s approach to workforce transformation, which Zavery described as beginning with cultural acceptance that AI is useful and transformative. At ServiceNow, this involved comprehensive employee retraining and providing access to AI capabilities so employees could experience firsthand how AI improves productivity.
Zavery emphasized that AI helps eliminate “skull-crushing” repetitive tasks, allowing employees to focus on more meaningful work. This approach has been implemented across ServiceNow’s engineering, finance, customer support, and go-to-market functions, with employees who initially feared displacement beginning to see AI as enabling higher-value activities.
Addressing concerns about AI-driven layoffs, Zavery suggested that some companies may be using AI as an excuse for workforce reductions rather than leveraging AI’s potential for business expansion. He noted that ServiceNow’s AI business growth allowed them to “add more people” and “expand our tap” into new market segments by reinvesting efficiency gains. Zavery provided historical context, noting that similar transformation anxiety “has happened with the cloud” and “with the web.”
Security Challenges in Agentic AI
Security emerged as the most significant barrier to enterprise AI adoption. Zavery reported that ServiceNow saw agentic workflow adoption increase by 55 times “late last year” once customers became comfortable with security and control mechanisms. This dramatic growth demonstrated that security concerns represented actual adoption barriers with quantifiable business impact.
The technical challenges of securing agentic AI systems are complex because these agents can change roles dynamically, sometimes every second, based on requirements. This necessitates sophisticated permission management that goes beyond traditional access control. ServiceNow’s acquisition of Vesa, which specializes in access graphs for non-human identities, illustrates the technical innovation required to address these challenges.
Zavery emphasized that security “cannot be on the side. It has to be part of the product” rather than being treated as an add-on component. ServiceNow’s security business, which Zavery noted is “billion dollars plus,” reflects the importance of integrated security approaches.
Enterprise Adoption Reality
The conversation provided a realistic assessment of enterprise AI adoption. Zavery acknowledged that early expectations for widespread agentic AI were overly optimistic, primarily due to missing foundational technologies around security and control. However, he noted that adoption has accelerated as these issues have been addressed.
Current adoption patterns show a shift from experimental implementations to production-focused deployments. Enterprises are now more willing to implement AI in critical business processes because they have necessary security and control mechanisms. Zavery observed that once enterprises successfully implement initial AI use cases and see return on investment, subsequent implementations accelerate rapidly.
AI Integration vs. Replacement in Software
Addressing predictions that 50% of enterprise software could shift to AI-based solutions, Zavery presented a nuanced perspective favoring integration over replacement. He argued that foundational AI models contribute only 5-10% of the intellectual property in enterprise solutions, while 90% of the value comes from contextual understanding and enterprise-specific functionality that software companies build around these models.
Zavery explained that software companies provide “the why part, the context part” while AI models tell “what to do, but they don’t know why.” This positions AI as a powerful component within software platforms rather than a replacement for them. The key differentiator for software companies will be their ability to understand specific domains, handle exceptions, and provide contextual reasoning around AI recommendations.
Future Directions: Physical AI
Looking forward, Zavery briefly identified operational technology and manufacturing as the next frontier for AI transformation, with humanoids and other physical AI systems beginning to change factory operations. He noted that this evolution from software-based AI to physical AI implementations introduces new categories of security and integration challenges.
Regulatory Considerations
The discussion touched on emerging regulatory frameworks, with Zavery noting that different countries are developing varying approaches to AI regulation, particularly around privacy and security. This regulatory development will likely influence enterprise AI adoption, especially in highly regulated industries.
Conclusion
The conversation revealed that successful enterprise AI adoption requires addressing technical, human, security, and governance challenges simultaneously. The dramatic adoption increases demonstrated by ServiceNow suggest enterprises are ready to embrace AI when fundamental concerns about trust, security, and control are adequately addressed.
Rather than replacing jobs or software, the discussion presented AI as an enabling technology requiring careful integration with existing systems and human capabilities. Zavery’s emphasis on trust as infrastructure provides a framework for understanding why some AI implementations succeed while others fail to gain enterprise traction. Companies that successfully navigate current trust and security challenges while building integrated AI capabilities will be best positioned for success in an AI-enabled business environment.
Session transcript
Ladies and gentlemen, and now I have the privilege of inviting our last speaker for the day, Mr. Amit Zavery, President and Chief Product Officer, ServiceNow. Mr. Zavery has spent his career at the intersection of enterprise software and AI, most recently leading ServiceNow’s push to embed AI agents into every corner of enterprise workflow. His perspective on agentic AI what it actually delivers versus what it promises is grounded in millions of enterprise deployments. He’ll be in conversation with Arjun Kharpal, CNBC’s Senior Tech Correspondent. Please welcome our guest and the moderator.
Hi, thanks so much for joining us. And if you’re watching online, thank you so much. Amit, let’s just kick off. You’ve got this view that trust is the new infrastructure in this age of AI. Can you just unpack what that means?
Yeah. Thank you, Arjun. I think if you look at what’s going on in the AI space, there’s a huge amount of interest in terms of using it in enterprise use cases as well, right? And without understanding what it’ll do for you and having any idea of what it landed up implementing inside your system, it becomes very hard to really depend on it. So that’s why without trust and safety and understanding of what’s happening in your underlying environment, it becomes very hard to expect to use AI in a lot of these enterprise use cases because your companies will not be able to do any auditing, compliance, visibility, and you wouldn’t really be able to really not run business without any kind of understanding of what’s going on.
So trust has to be a big part of it.
And trust, I guess, in the enterprise sense of the word probably has lots of different definitions, right? We’re talking about trust amongst employees, for example, but also from the cyber perspective, from the security and safety perspective you were just mentioning there. So it’s worth digging into some of these. How do you design some of these? In the enterprise, let’s start with perhaps the human element. at this point, because there’s a lot of concern from people right now about potential job losses and the impact AI could have on their roles as well. So from the human perspective, how do you design trust within the organization?
Yeah, no, I think it’s still something which I think the industry is still trying to figure out, to be honest. The way to think through this one is, one, everybody has to accept that AI is useful, and there’s a lot of opportunity to embed that in terms of your day -to -day lives. Second, I think there is a reality that this thing is transforming how the world works, and it’s moving very fast. So once you understand the principles and the value of it, then you start building together in terms of what the cultural shifts need to be, how people need to work together, and how do you help them understand the value while keeping their jobs very important and be able to bring them into the conversation.
So there’s a huge amount of cultural shift inside the company, as well as being able to kind of educate. Everybody in terms of what it delivers for you. So what we’ve been doing at ServiceNow, for example, we’ve been retraining our employees and giving them access to a lot of the AI capabilities, making sure they get to see what it does for their day -to -day life at the employee productive level and see that, okay, you know what, I could do my job faster, better, more efficiently, and free up more of my time to do other things which I couldn’t get to. Second thing after that, once you get them re -skilled, is to really now take it to the enterprise level, not just an employee.
Like, now how do you improve your processes? And the processes which are cutting across multiple departments, can I make that thing work faster? Can I get a better understanding of how it operates? And can you land up freeing up a lot of the human painful work you used to do, right? A lot of the repetitive tasks, which we used to call a skull -crushing task, which is making it so difficult in the day -to -day life that they can’t really get anything done beyond that. So if you remove those barriers, people start trusting that, oh, you know what, this is helping me. It’s getting my job done better. And it’s also getting me more understanding of new technologies.
So, I’m going to go ahead and start talking about the technology that I’m using. And you start accepting that in your day -to -day work environment and take that to the next level because you start innovating. So it’s a step process, I would say. And what we have done today at ServiceNow, we’ve seen the adoption go up a lot, be it in engineering, be it in finance, be it in customer support, be it in go -to -market, because they started playing with these technologies and bringing it to the day -to -day work environment. And from there, they’ve been starting to now innovate and come up with new ideas to help make their jobs better and how you make the customer’s life better long -term.
Does that set them up then for success when, you know, inevitably we will see the changing nature of work? And also, we’ve already seen some companies, you know, make layoffs and blame it on AI, whether that’s true or not is another debate. But certainly, there will be changing nature of work and organizations are rethinking the workforce. So by doing the reskilling, is this… Setting employees up.
No, I think you’re right. I mean, with any technology transformation, there’s always worries about job losses. It’s nothing, this is not the first time a technology shift has created that anxiety. It has happened with the cloud when the cloud happened. It happened with the web when the transformation happened towards that. I think the difference here is the speed sometimes and the uncertainty of understanding what it does to you as an individual in some cases. But I think the worry, a lot of the news out there is saying because of AI, we reduce our staffing. I think some of them are just using that as an excuse I’ve seen so far. If you look at our business, our AI business has grown significantly.
And we have been able to add more people because we were able to expand our tap. We’ve been able to get into a lot more new segments of market because of the investment. We’ve been able to reinvest a lot of that money we save because of AI into a lot of new areas. And that is one thing which I think a lot of companies are starting to realize. You take out a lot of the mundane tasks and move into the high value tasks. You can increase your top. line. Sure, you help on the bottom line with AI. There is a lot of work you can now outsource to autonomous agents and agentic workflows instead of having to do it by humans.
But the humans are now able to do a lot of other things you couldn’t and get into a lot of new segments. Now, business has grown significantly because we’ve been able to now take that savings and invest into a lot of new areas.
You mentioned agentic there, and I’m glad you did because the other part of this trust equation is what you were mentioning earlier around safety and security as well. Well, given the excitement around agentic AI and how much businesses want to adopt this, is there enough focus being put right now on the vulnerabilities from a cyber perspective when it comes to agentic AI?
I would say that’s probably the biggest concern for companies when they think about AI and agentic, right? If you look at last year, early last year, when we used to go and talk about agentic workflows or AI, most of the companies were worried about not having any kind of visibility, worried about vulnerabilities, worried about security, worried about control. And once we started introducing them the capability of controlling some of this implementation and having security profiles around the AI implementation, a lot more companies started adopting AI. Late last year, I would say middle of last year and late last year, the volume of our agentic workflows being adopted by customers went up by 55 times, 55x.
Because what happened was they started feeling comfortable that one, they have visibility into all the AI systems. Second, they have ability to secure it because you don’t want to lose access to your data or get it accessed externally without any kind of permissions. And once you start giving them that comfort factor, they’re starting to see the benefit of taking agentic AI and implementing that into the businesses, be a workflow around case management, incident management, triaging, be able to resolve issues. And that is a very, very valuable things for them. But once you only can do that once you have the security part of it. So, for example, we’ve been. Investing aggressively in the security space, our security.
business itself is a billion dollars plus, but we’ve been adding now for AI agents. The AI agents are changing roles every second you call them based on what requirements you have. So how do you manage the permissions? How do you manage their identity? So we bought recently a company called Vesa, which does access graphs for non -human identities, which makes it much more valuable to our customers because now they know that those agents are guaranteed to not do something nefarious or they won’t have access to data you’re not allowed to have. And whenever you change the roles, they’re only getting to do things based on the roles. So it’s a very important part of it.
And I think agentic AI and things like that will not be adopted if you don’t have a right kind of security technology as part of this implementation. It cannot be on the side. It has to be part of the product.
There’s certain tech companies who will sort of talk up the capabilities of agentic AI right now and talk up how enterprises are adopting AI. But what do you think? I think it’s a very important part of it. I think it’s a very important part of it. From your perspective, has the adoption of AI from enterprises been faster, slower, or about right than you had anticipated?
I think there was a lot of expectation early last year. Everybody thought that agentic AI and AI agents will be proliferated across every enterprise. I thought that was probably a little more optimistic and unrealistic because there were a lot of technologies which are missing to really provide you a platform which guarantees everything before you go and adopt it. A lot of those things started happening, I think I would say middle last year, and now the volume of adoption has gone up. But it is probably more thoughtful than probably experimental the way it was before. A lot of people were experimenting with it, but they were not wanting to put it in production because of the security things you talked about, trust and safety and compliance.
Now with a lot of the things customers are seeing from vendors like us, where you’re providing AI control tower, for example, to make sure you have visibility and control, they’re feeling more comfortable. So the volume is starting to go up. Use cases are getting much more defined. And what I’ve seen so far is that once you implement one or two use cases, you start seeing ROI. then the next more use cases become very very fast so you have to make it easy to be adopted you have to provide the security and everything else around it and then get them to see the roi and once you get the roi i think the customers all feel that this is something valuable to them and it’s something they want to invest in
I mean can i get your take on a on a comment we had on cnbc this week i was speaking to the ceo of mistral ai in europe and it was around this conversation happening in financial markets right now around software yeah um and how much are these agentic ai systems going to do the job of software that enterprises currently pay for uh and these sas businesses and he said you know he believes that 50 roughly of current you know software being used by enterprises uh could shift to ai i just wanted to get your take on that given how how embedded you are in this industry
No, I think that there’s there’s a lot of people who are in the industry who are in the industry who are in the industry a lot of this debate about what is ai going to do the software industry i think uh ai is going to be a synergetic part of Any software you’re going to build going forward, and it’s already happening now, has to be with AI mindset and AI as part of the platform and the foundation. The companies which are going to suffer are the companies who are not adopting AI fast enough. So any vendor who’s thinking about AI as a side thing or something which is coming later, I think it’ll be very difficult to really justify customers buying that product.
Companies like us and others who are starting to make, we have been doing that for a few years, where they’ve been making AI part of the foundation, part of the platform. We’re already accelerating that adoption because customers, once they value, second, I think they do believe that this is going to be a very competitive advantage to them as well. And so we see a lot of synergy. We do a lot of partnership with OpenAI, Anthropic. We work with Mistral. We work with Google and Gemini because I think there’s a synergy between what foundational models and AI technology provides. and all the things you have to do around it. That’s what software industry can do. So what we’re doing is we’re building on top of it, but it’s like 5 % to 10 % of IP comes from those models.
90 % comes from technology we build because you have to build a lot of context around enterprise use cases. You have to understand what it means. You have to understand why an exception happened, how you handle it. Models are basically telling them what to do, but they don’t know why. The why part, the context part, comes from technologies and software we build. And the companies who are going to do that much more better, understand domain, understand expertise, and have a lot of experience, will win in this market. And that’s the difference, I think.
I mean, we’ve got about a minute left. I just wanted to get your take on the future. If we were sat here
I think we still will be talking about security and risk, definitely, because there’s a lot of work still to be done. regulations. Every country is now thinking about what AI means to them, what kind of regulations they want to put in for privacy, security, other things like that. I think the other one which is starting to come up a lot is physical AI. So we’re doing a lot of work in OT, operational technology, because a lot of the shop factories are changing with physical AI, with humanoids and droids and things like that, because they are going to be the next generational way of manufacturing. So how do you now secure that? How do you bring that as part of the processes?
How you integrate that into your environment is going to be a critical discussion as well.
Fantastic. Amit, thanks for your insights. So incisive. I appreciate your time. Thank you so much. Round of applause for Amit Zaveri of ServiceCamp. Thank you, everyone.
Mr. Amit Zaveri, and thanks, Arjun Kharpal, for moderating this conversation. Ladies and gentlemen, with this, we end.
Amit Zavery
Speech speed
218 words per minute
Speech length
2134 words
Speech time
586 seconds
Trust necessity for AI adoption
Explanation
Amit stresses that without trust, safety, and clear visibility into AI systems, enterprises cannot audit, comply, or run their business effectively, making trust a prerequisite for AI deployment.
Evidence
“So that’s why without trust and safety and understanding of what’s happening in your underlying environment, it becomes very hard to expect to use AI in a lot of these enterprise use cases because your companies will not be able to do any auditing, compliance, visibility, and you wouldn’t really be able to really not run business without any kind of understanding of what’s going on.” [4]. “So trust has to be a big part of it.” [2].
Major discussion point
Trust as the foundational infrastructure for enterprise AI
Topics
Building confidence and security in the use of ICTs | Artificial intelligence
Cultural shift, reskilling, and productivity gains
Explanation
Amit describes a large cultural shift inside companies, emphasizing education and reskilling of staff so AI can be leveraged at the enterprise level, leading to business growth and reinvestment of savings.
Evidence
“So there’s a huge amount of cultural shift inside the company, as well as being able to kind of educate.” [21]. “Second thing after that, once you get them re -skilled, is to really now take it to the enterprise level, not just an employee.” [23]. “Now, business has grown significantly because we’ve been able to now take that savings and invest into a lot of new areas.” [33]. “So what we’ve been doing at ServiceNow, for example, we’ve been retraining our employees and giving them access to a lot of the AI capabilities, making sure they get to see what it does for their day -to -day life at the employee productive level and see that, okay, you know what, I could do my job faster, better, more efficiently, and free up more of my time to do other things which I couldn’t get to.” [44].
Major discussion point
Designing trust among employees and managing workforce impact
Topics
Capacity development | Building confidence and security in the use of ICTs
AI does not inherently cause layoffs; enables higher‑value work
Explanation
Amit argues that AI removes mundane tasks, allowing employees to focus on higher‑value activities, and that the technology can improve the bottom line without necessarily reducing headcount.
Evidence
“You take out a lot of the mundane tasks and move into the high value tasks.” [29]. “So what we’ve been doing at ServiceNow, for example, we’ve been retraining our employees … and free up more of my time to do other things which I couldn’t get to.” [44].
Major discussion point
Designing trust among employees and managing workforce impact
Topics
Capacity development | The digital economy
Visibility, vulnerability, and control concerns; security profiles boost confidence
Explanation
Amit notes that early concerns centered on lack of visibility, vulnerabilities, and control, but providing AI control‑tower and security profiles has increased enterprise comfort and adoption.
Evidence
“If you look at last year, early last year, when we used to go and talk about agentic workflows or AI, most of the companies were worried about not having any kind of visibility, worried about vulnerabilities, worried about security, worried about control.” [51]. “Now with a lot of the things customers are seeing from vendors like us, where you’re providing AI control tower, for example, to make sure you have visibility and control, they’re feeling more comfortable.” [52].
Major discussion point
Security, safety, and compliance for agentic AI
Topics
Building confidence and security in the use of ICTs | Artificial intelligence
Acquisition of Vesa for non‑human identity access
Explanation
Amit explains that ServiceNow acquired Vesa, which provides access graphs for non‑human identities, helping customers ensure agents cannot act maliciously or access unauthorized data.
Evidence
“So we bought recently a company called Vesa, which does access graphs for non -human identities, which makes it much more valuable to our customers because now they know that those agents are guaranteed to not do something nefarious or they won’t have access to data you’re not allowed to have.” [59].
Major discussion point
Security, safety, and compliance for agentic AI
Topics
Building confidence and security in the use of ICTs | Artificial intelligence
Early optimism was unrealistic; adoption now more thoughtful
Explanation
Amit reflects that early expectations were overly optimistic given missing technologies, and that current adoption is more measured and focused on trust and security foundations.
Evidence
“I thought that was probably a little more optimistic and unrealistic because there were a lot of technologies which are missing to really provide you a platform which guarantees everything before you go and adopt it.” [73]. “But it is probably more thoughtful than probably experimental the way it was before.” [75]. “I think there was a lot of expectation early last year.” [76].
Major discussion point
Pace and expectations of enterprise AI adoption
Topics
Artificial intelligence | The digital economy
AI as a synergistic layer, not a wholesale replacement
Explanation
Amit asserts that AI will become a synergistic component embedded in every software platform, with domain expertise and context still essential for success.
Evidence
“No, I think that there’s there’s a lot of people who are in the industry who are in the industry who are in the industry a lot of this debate about what is ai going to do the software industry i think uh ai is going to be a synergetic part of Any software you’re going to build going forward, and it’s already happening now, has to be with AI mindset and AI as part of the platform and the foundation.” [45]. “The companies who are going to do that much more better, understand domain, understand expertise, and have a lot of experience, will win in this market.” [84].
Major discussion point
AI’s impact on the software industry and future of SaaS
Topics
Artificial intelligence | The digital economy
Physical AI/OT and ongoing security/regulatory needs
Explanation
Amit highlights work in operational technology with physical AI (humanoids, droids) and stresses that security, risk, and regulation will remain critical as AI expands into manufacturing.
Evidence
“So we’re doing a lot of work in OT, operational technology, because a lot of the shop factories are changing with physical AI, with humanoids and droids and things like that, because they are going to be the next generational way of manufacturing.” [48]. “I think we still will be talking about security and risk, definitely, because there’s a lot of work still to be done.” [54]. “Every country is now thinking about what AI means to them, what kind of regulations they want to put in for privacy, security, other things like that.” [67].
Major discussion point
Future outlook: regulation, risk, and physical AI/OT
Topics
Building confidence and security in the use of ICTs | Artificial intelligence | The enabling environment for digital development
Arjun Kharpal
Speech speed
190 words per minute
Speech length
516 words
Speech time
162 seconds
Multi‑dimensional nature of trust in enterprises
Explanation
Arjun frames trust as the new infrastructure for AI and notes that trust in an enterprise context can have many definitions, underscoring its multi‑faceted nature.
Evidence
“You’ve got this view that trust is the new infrastructure in this age of AI.” [1]. “And trust, I guess, in the enterprise sense of the word probably has lots of different definitions, right?” [16].
Major discussion point
Trust as the foundational infrastructure for enterprise AI
Topics
Building confidence and security in the use of ICTs | Artificial intelligence
Preparing employees for changing work (reskilling)
Explanation
Arjun asks whether reskilling efforts are setting employees up for success as the nature of work evolves with AI.
Evidence
“So by doing the reskilling, is this…” [24]. “Does that set them up then for success when, you know, inevitably we will see the changing nature of work?” [28].
Major discussion point
Designing trust among employees and managing workforce impact
Topics
Capacity development | Building confidence and security in the use of ICTs
Inquiry about cyber‑security focus for agentic AI
Explanation
Arjun questions whether enough attention is being given to cyber‑security vulnerabilities as businesses rush to adopt agentic AI.
Evidence
“Well, given the excitement around agentic AI and how much businesses want to adopt this, is there enough focus being put right now on the vulnerabilities from a cyber perspective when it comes to agentic AI?” [58].
Major discussion point
Security, safety, and compliance for agentic AI
Topics
Building confidence and security in the use of ICTs
Question on adoption speed (faster/slower/about right)
Explanation
Arjun probes Amit on whether enterprise AI adoption has been faster, slower, or on target compared with expectations.
Evidence
“From your perspective, has the adoption of AI from enterprises been faster, slower, or about right than you had anticipated?” [3].
Major discussion point
Pace and expectations of enterprise AI adoption
Topics
Artificial intelligence | The digital economy
Claim that up to 50 % of enterprise software could shift to AI
Explanation
Arjun cites a comment from a CNBC interview suggesting that roughly half of current enterprise software could be replaced by AI, and asks Amit for his take.
Evidence
“…he said you know he believes that 50 roughly of current you know software being used by enterprises uh could shift to ai…” [86].
Major discussion point
AI’s impact on the software industry and future of SaaS
Topics
Artificial intelligence | The digital economy
Closing prompt for future perspective
Explanation
Arjun ends the discussion by inviting Amit to share his outlook on the future of AI and enterprise technology.
Evidence
“I just wanted to get your take on the future.” [70].
Major discussion point
Future outlook: regulation, risk, and physical AI/OT
Topics
Artificial intelligence | The enabling environment for digital development
Moderator
Speech speed
123 words per minute
Speech length
111 words
Speech time
53 seconds
Establishing speaker authority on AI and enterprise software
Explanation
The moderator highlights Amit Zavery’s extensive background at the intersection of enterprise software and AI, and his leadership in embedding AI agents throughout ServiceNow’s workflow, thereby setting a credible foundation for the discussion.
Evidence
“Mr. Zavery has spent his career at the intersection of enterprise software and AI, most recently leading ServiceNow’s push to embed AI agents into every corner of enterprise workflow.” [6]. “Ladies and gentlemen, and now I have the privilege of inviting our last speaker for the day, Mr. Amit Zavery, President and Chief Product Officer, ServiceNow.” [4].
Major discussion point
Trust as the foundational infrastructure for enterprise AI
Topics
Artificial intelligence | The digital economy
Framing the gap between AI promises and actual delivery
Explanation
The moderator signals that the conversation will examine what agentic AI truly delivers versus the expectations set by its hype, prompting a critical assessment of AI capabilities.
Evidence
“His perspective on agentic AI what it actually delivers versus what it promises is grounded in millions of enterprise deployments.” [7].
Major discussion point
AI’s impact on the software industry and future of SaaS
Topics
Artificial intelligence | The digital economy
Signaling the strategic importance of AI integration across enterprise workflows
Explanation
By introducing Zavery as the final speaker and noting his role in spreading AI agents throughout enterprise processes, the moderator underscores AI’s central role in modern business operations.
Evidence
“Ladies and gentlemen, and now I have the privilege of inviting our last speaker for the day, Mr. Amit Zavery, President and Chief Product Officer, ServiceNow.” [4]. “Mr. Zavery has spent his career at the intersection of enterprise software and AI, most recently leading ServiceNow’s push to embed AI agents into every corner of enterprise workflow.” [6].
Major discussion point
AI as a synergistic layer, not a wholesale replacement
Topics
Artificial intelligence | The digital economy
Concluding the dialogue and emphasizing ongoing engagement
Explanation
The moderator formally ends the session, indicating that the discussion has covered key AI themes and that continued conversation will be essential for future developments.
Evidence
“Ladies and gentlemen, with this, we end.” [5].
Major discussion point
Future outlook: regulation, risk, and physical AI/OT
Topics
Artificial intelligence | The enabling environment for digital development
Agreements
Agreement points
Trust is fundamental for enterprise AI adoption
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
Trust is the new infrastructure in AI age, requiring understanding and visibility for enterprise adoption
Trust encompasses multiple dimensions including human, security, and safety perspectives in enterprise contexts
Summary
Both speakers agree that trust forms the foundational requirement for successful enterprise AI implementation, though they emphasize different dimensions – Zavery focuses on operational trust while Kharpal emphasizes the multi-faceted nature of trust including human and security elements
Topics
Artificial intelligence | Building confidence and security in the use of ICTs
Security and visibility are critical barriers to AI adoption
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
Without trust, safety, and understanding, enterprises cannot conduct auditing, compliance, and visibility necessary for business operations
Security vulnerabilities and lack of visibility are major concerns preventing enterprise adoption of agentic AI
Summary
Both speakers recognize that security concerns and lack of visibility into AI operations are primary obstacles preventing enterprises from adopting AI systems at scale
Topics
Building confidence and security in the use of ICTs | Artificial intelligence
Employee preparation and reskilling are essential for AI transformation
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
Cultural shift and employee education are essential, requiring retraining and demonstrating AI value in daily work
Reskilling employees prepares them for the changing nature of work and potential organizational restructuring
Summary
Both speakers acknowledge that successful AI adoption requires comprehensive employee preparation, including reskilling and cultural adaptation to new ways of working
Topics
Capacity development | Artificial intelligence | The digital economy
AI adoption patterns show acceleration after initial success
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
Current adoption is more thoughtful and production-focused rather than experimental, driven by improved security and control capabilities
Enterprise adoption patterns show accelerating use case implementation once initial ROI is demonstrated
Summary
Both speakers observe that enterprise AI adoption follows a pattern where initial cautious implementation leads to rapid scaling once benefits and security are demonstrated
Topics
Artificial intelligence | The digital economy
Similar viewpoints
Both speakers recognize that AI systems, whether digital agents or physical AI, require sophisticated new security frameworks that go beyond traditional approaches due to their dynamic and autonomous nature
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
AI agents require dynamic permission management and identity control as they change roles frequently
Integration of humanoids and droids into enterprise processes will require new security and process frameworks
Topics
Building confidence and security in the use of ICTs | Artificial intelligence
Both speakers share a realistic assessment of AI adoption timelines, acknowledging that initial expectations were too optimistic but that adoption is now accelerating as foundational issues are addressed
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
Early expectations for agentic AI proliferation were overly optimistic due to missing foundational technologies
Enterprise adoption patterns show accelerating use case implementation once initial ROI is demonstrated
Topics
Artificial intelligence | The enabling environment for digital development
Unexpected consensus
AI as business growth driver rather than job destroyer
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
AI can drive business growth and job creation by expanding market reach and enabling investment in new areas
Reskilling employees prepares them for the changing nature of work and potential organizational restructuring
Explanation
While job displacement from AI is a common concern, both speakers converge on the view that AI can actually drive job creation and business expansion when properly implemented with appropriate workforce preparation
Topics
The digital economy | Artificial intelligence | Social and economic development
Security must be built-in rather than added-on
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
Security must be integrated into the product, not treated as a separate component
Security vulnerabilities and lack of visibility are major concerns preventing enterprise adoption of agentic AI
Explanation
Both speakers unexpectedly align on the critical insight that security cannot be an afterthought in AI systems but must be architected into the core product design from the beginning
Topics
Building confidence and security in the use of ICTs | Artificial intelligence
Overall assessment
Summary
The speakers demonstrate strong consensus on fundamental issues around enterprise AI adoption, particularly regarding the critical importance of trust, security, and workforce preparation. They agree on realistic timelines for adoption and the need for integrated security approaches.
Consensus level
High level of consensus with complementary perspectives – Zavery provides the vendor/implementation perspective while Kharpal offers the industry analyst viewpoint, but both arrive at similar conclusions about the key challenges and requirements for successful enterprise AI adoption. This consensus suggests these are well-established industry insights rather than controversial positions.
Differences
Different viewpoints
Impact of AI on software industry transformation
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
AI must be foundational to software platforms rather than a side feature for competitive advantage
Significant portion of current enterprise software could potentially shift to AI-based solutions
Summary
Kharpal presents the view that 50% of enterprise software could shift to AI (citing Mistral AI CEO), while Zavery argues that AI should be integrated into existing software platforms rather than replacing them, emphasizing that software companies provide 90% of value through context and domain expertise
Topics
Artificial intelligence | The digital economy
Unexpected differences
Overall assessment
Summary
The discussion shows minimal disagreement, with only one clear disagreement on software industry transformation and two areas of partial agreement on workforce transformation and security approaches
Disagreement level
Low disagreement level – this was primarily a collaborative interview where Kharpal asked probing questions and Zavery provided detailed responses. The disagreements were more about different perspectives on the same issues rather than fundamental opposition, which is typical for an interview format rather than a debate
Partial agreements
Partial agreements
Both agree that AI will transform work and that employee preparation is important, but they approach it differently – Zavery focuses on demonstrating AI value through task automation and reskilling, while Kharpal raises concerns about whether reskilling adequately prepares workers for potential job displacement and organizational restructuring
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
AI adoption should focus on removing repetitive “skull-crushing” tasks while enabling employees to do higher-value work
Reskilling employees prepares them for the changing nature of work and potential organizational restructuring
Topics
The digital economy | Capacity development | Artificial intelligence
Both recognize security as a critical barrier to AI adoption, but Zavery emphasizes the solution of integrating security into products and demonstrates success with this approach, while Kharpal questions whether the industry is doing enough to address these vulnerabilities
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
Security must be integrated into the product, not treated as a separate component
Security vulnerabilities and lack of visibility are major concerns preventing enterprise adoption of agentic AI
Topics
Building confidence and security in the use of ICTs | Artificial intelligence
Similar viewpoints
Both speakers recognize that AI systems, whether digital agents or physical AI, require sophisticated new security frameworks that go beyond traditional approaches due to their dynamic and autonomous nature
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
AI agents require dynamic permission management and identity control as they change roles frequently
Integration of humanoids and droids into enterprise processes will require new security and process frameworks
Topics
Building confidence and security in the use of ICTs | Artificial intelligence
Both speakers share a realistic assessment of AI adoption timelines, acknowledging that initial expectations were too optimistic but that adoption is now accelerating as foundational issues are addressed
Speakers
– Amit Zavery
– Arjun Kharpal
Arguments
Early expectations for agentic AI proliferation were overly optimistic due to missing foundational technologies
Enterprise adoption patterns show accelerating use case implementation once initial ROI is demonstrated
Topics
Artificial intelligence | The enabling environment for digital development
Takeaways
Key takeaways
Trust is the foundational infrastructure for enterprise AI adoption, requiring visibility, understanding, and control mechanisms for auditing and compliance
Successful AI implementation requires a step-by-step cultural transformation: employee education and reskilling, demonstrating value in daily work, then scaling to enterprise-level processes
Security and safety concerns are the primary barriers to agentic AI adoption, but providing proper security profiles and control mechanisms can dramatically increase adoption (55x growth demonstrated)
AI adoption has shifted from experimental to thoughtful production deployment, with companies seeing accelerated implementation once initial ROI is proven
AI must be foundational to software platforms rather than an add-on feature, with software companies providing 90% of value through enterprise context while foundational models contribute only 5-10%
AI can drive business growth and job creation by removing repetitive tasks, enabling higher-value work, and allowing companies to expand into new market segments
Resolutions and action items
ServiceNow has implemented AI control towers to provide visibility and control for enterprise customers
ServiceNow acquired Vesa company to manage access graphs for non-human identities and dynamic permission management for AI agents
ServiceNow continues partnerships with OpenAI, Anthropic, Mistral, and Google to integrate foundational models with enterprise context
Companies should focus on reskilling employees and demonstrating AI value in daily work before scaling to enterprise processes
Unresolved issues
The industry is still figuring out how to design trust within organizations from the human perspective
Regulatory frameworks for AI are still being developed by different countries with varying approaches to privacy and security
Integration of physical AI, humanoids, and droids into enterprise processes requires new security and process frameworks that are still being developed
The debate continues about what percentage of current enterprise software will be replaced by AI solutions
Suggested compromises
Take a gradual approach to AI implementation: start with employee-level productivity, then move to enterprise-level processes
Balance AI automation with human workforce by focusing AI on repetitive tasks while enabling humans to do higher-value work
Integrate security as part of the AI product rather than treating it as a separate component
Combine foundational AI models with enterprise-specific context and domain expertise rather than relying solely on either approach
Thought provoking comments
Trust is the new infrastructure in this age of AI… without trust and safety and understanding of what’s happening in your underlying environment, it becomes very hard to expect to use AI in a lot of these enterprise use cases because your companies will not be able to do any auditing, compliance, visibility
Speaker
Amit Zavery
Reason
This reframes trust from a soft concept to a foundational technical requirement, positioning it as critical infrastructure rather than just a nice-to-have. It connects abstract concerns about AI adoption to concrete business needs like auditing and compliance.
Impact
This opening statement set the entire framework for the discussion, establishing trust as the central theme that would be explored from multiple angles – human, security, and operational perspectives throughout the conversation.
I think some of them are just using that as an excuse I’ve seen so far… You take out a lot of the mundane tasks and move into the high value tasks. You can increase your top line… The humans are now able to do a lot of other things you couldn’t and get into a lot of new segments.
Speaker
Amit Zavery
Reason
This challenges the dominant narrative about AI-driven job losses by suggesting companies may be using AI as a convenient excuse for layoffs while presenting a counter-narrative of AI as a business expansion tool rather than just a cost-cutting measure.
Impact
This shifted the conversation from defensive concerns about job displacement to a more optimistic view of AI as enabling business growth and human potential, providing a different lens through which to view workforce transformation.
The volume of our agentic workflows being adopted by customers went up by 55 times, 55x. Because what happened was they started feeling comfortable that one, they have visibility into all the AI systems. Second, they have ability to secure it
Speaker
Amit Zavery
Reason
This provides concrete evidence that security and visibility aren’t just theoretical concerns but actual adoption barriers, with dramatic quantitative proof (55x increase) of what happens when these concerns are addressed.
Impact
This data point validated the earlier trust-as-infrastructure thesis with hard numbers, making the abstract concept tangible and demonstrating the business impact of addressing trust concerns.
AI is going to be a synergetic part of any software you’re going to build going forward… The companies which are going to suffer are the companies who are not adopting AI fast enough… 5% to 10% of IP comes from those models. 90% comes from technology we build
Speaker
Amit Zavery
Reason
This challenges the notion that AI will replace software by arguing instead that AI becomes embedded within software, and provides a specific breakdown of value creation that counters fears about foundational models dominating the software industry.
Impact
This comment directly addressed the interviewer’s question about AI potentially replacing 50% of enterprise software, reframing the discussion from replacement to integration and emphasizing the continued importance of domain expertise and context.
I think the other one which is starting to come up a lot is physical AI. So we’re doing a lot of work in OT, operational technology, because a lot of the shop factories are changing with physical AI, with humanoids and droids
Speaker
Amit Zavery
Reason
This introduces a forward-looking perspective that extends beyond current software-focused AI discussions to physical manifestations, suggesting the next wave of AI transformation will be in manufacturing and operational technology.
Impact
This comment opened up a new dimension to the conversation, expanding the scope from current enterprise software AI to future physical AI applications, suggesting the discussion topics will evolve significantly in coming years.
Overall assessment
These key comments shaped the discussion by establishing a comprehensive framework that moved beyond typical AI hype to address practical enterprise concerns. Zavery’s opening trust-as-infrastructure concept provided the foundation, while his subsequent comments systematically addressed common fears (job losses, security vulnerabilities, software replacement) with data-driven counter-narratives and practical examples. The conversation evolved from defensive concerns to growth opportunities, culminating in forward-looking insights about physical AI. Rather than a typical promotional interview, this became a nuanced exploration of AI adoption barriers and solutions, grounded in real enterprise experience and quantifiable outcomes.
Follow-up questions
How do companies effectively manage permissions and identity for AI agents that change roles dynamically?
Speaker
Amit Zavery
Explanation
This is a critical technical challenge as AI agents adapt their roles every second based on requirements, requiring sophisticated access management systems
What specific regulations will different countries implement for AI privacy and security?
Speaker
Amit Zavery
Explanation
Every country is developing AI regulations, but the specific frameworks and their impact on enterprise AI adoption remain unclear
How can physical AI be integrated into existing enterprise environments and processes?
Speaker
Amit Zavery
Explanation
Physical AI including humanoids and droids in manufacturing represents the next generation of operational technology that needs integration with current systems
How do you secure operational technology environments that incorporate physical AI?
Speaker
Amit Zavery
Explanation
As factories adopt physical AI technologies, new security frameworks are needed to protect these operational technology environments
What are the best practices for cultural transformation when implementing AI in enterprises?
Speaker
Amit Zavery
Explanation
The industry is still figuring out how to manage the cultural shifts required for successful AI adoption while maintaining employee trust and engagement
Disclaimer: This is not an official session record. DiploAI generates these resources from audiovisual recordings, and they are presented as-is, including potential errors. Due to logistical challenges, such as discrepancies in audio/video or transcripts, names may be misspelled. We strive for accuracy to the best of our ability.
Related event
